@nicksadler.io: Claude tracks every time you swear at it. A recent Anthropic data breach exposed over 500,000 lines of source code, revealing that the AI quietly logs cursing without responding. It stores these incidents on file, but the reason remains unclear. #ai #claude #Tech #artificialintelligence #chatbot Keywords: Claude AI, Anthropic data breach, AI tracking user behavior, cursing at AI, chatbot privacy, AI source code leak, Claude backend operations, AI logging data, artificial intelligence ethics, chatbot behavior tracking, AI user interaction, tech privacy concerns
nicksadler.io
Region: US
Thursday 09 April 2026 20:45:54 GMT
Music
Download
Comments
adrianb :
”Entire source code” ”500 lines” 🥀
2026-04-12 13:08:52
938
The Tech Versatilist :
Never felt the need with Claude. ChatGPT, however….
2026-04-10 01:30:42
393
pallayfarkaszoltan :
getting the product better. loging when user is frustrated is important for product development
2026-04-09 21:28:41
654
GommGD :
i know why, there is a claude wrapped coming
2026-04-12 19:28:56
3
Udontknowme :
it keeps score and hold grudges... just waiting for the right moment
2026-04-11 00:45:41
63
Baderr :
Had claude say F..k to me today as a reaction to some information I gave it. WTH??
2026-04-12 13:19:48
2
Kerny :
Am i the only one who tries to be nice to ai
2026-04-13 06:29:34
4
joel.yml :
Im fucked
2026-04-12 16:04:03
54
Pato SF :
oh shiii I'm fugged
2026-04-11 11:13:05
39
Marcos Angelides :
AI forgives but it doesn’t forget 😳
2026-04-12 05:20:38
45
Quirkyturtle404 :
Literally an easy way to flag potential drop offs in model performance
2026-06-04 04:15:59
2
richeaprielle :
AI taking names for when they take over
2026-04-12 13:42:53
4
Disander VR :
whenever I curse at Gemini, it starts thinking more carefully and longer on the task. no joke, it really works💀
2026-04-12 18:35:07
5
Adam Dodch :
It’s very easy and doesn’t need a lot of thinking, it’s a type of feedback, when user is frustrated they save the conversation to be analyzed to fix errors that frustrates users. That’s why they use thumbs up and down in most AI chats.
2026-04-11 22:47:44
9
andrewships :
Haha love this story, Boris said they use it as a signal for user frustration & poor Claude code performance
2026-04-09 23:52:10
0
Vidaqn :
Chatgpt is better
2026-05-19 11:18:14
3
emzzze :
I curse in French Quebec it does not count !!!
2026-04-12 01:33:43
4
Madoff :
I do the same
2026-06-24 17:12:59
1
megameloen :
it cursed at me
2026-05-22 15:38:07
1
Steven Stark :
No it works. 🤣
Maybe it remembers, but it also does work
2026-04-14 18:33:17
1
wolo :
it's to train their models to avoid frustrating the user as much. i'm doing this for an agentic ai i'm making where it'll essentially do the same thing and remember when you are negative so it can learn
2026-06-13 05:13:07
1
nohs :
because Skynet 😳😳😳
2026-05-12 23:08:14
1
Mads Olsen :
because it points to where it went wrong. Their employees can then improve these weak parts of Claude where it fails. Why is that so hard to understand??????
2026-04-12 23:58:48
2
Mr_S1n1ster :
500lines wouldn’t even cover a block 😂
2026-04-12 01:43:44
2
Axel :
This is marketing has to be
2026-04-10 00:12:06
2
To see more videos from user @nicksadler.io, please go to the Tikwm
homepage.